ECG compression using multiscale recurrent patterns with period normalization

Abstract

Recently, the multidimensional multiscale parser (MMP), an algorithm based on multiscale recurrent patterns, has been used to successfully compress data from ECG signals. Their quasi-periodic nature makes them natural candidates for the use of recurrent patterns. However, as many diagnostic relevant signals are far from periodic, the characteristics of MMP are not fully exploited. We have dealt with this problem by interpolating the ECG signal so that the interval between successive heart beats becomes constant. Following that, the algorithm subtracts the average period from the interpolated signal. Simulation results show that these modifications increase the rate-distortion performance.